PDF to CSV your accounting software actually imports

Convert the PDF first

Keep the format on “Excel (.xlsx)” and set Sheets to “All on one” for a statement with several pages. The tidy pass below is easier in a sheet than in a text file.

The converter can also write CSV directly (Format, CSV, then a Separator). That file is fine for a quick look or a simple 1-page list, but it contains every line of the page. Our invented 1-page invoice gave a CSV of 17 lines, of which 4 were line items: the rest were the letterhead, the invoice number and date, a wrapped description, the totals and the bank details. An accounting import would read every one of those as a record.

Why not straight from PDF to CSV?

A CSV for an import has to be exactly 1 header row and then 1 record per row. A PDF page has a title, an address, headings, totals, notes and page numbers around its table. No converter can tell for certain which lines your software wants, so either it guesses and sometimes drops a real record, or it keeps everything and you delete the rest. Ours keeps everything, so nothing is lost without your knowing.

What accounting imports expect

Every import page differs, but they ask the same questions. Find them on your software’s import page and note the answers:

The tidy pass in Excel

1. Delete everything that is not a record. Title rows, the account holder, opening and closing balances, totals, page footers, the empty rows between pages, and the repeated heading row on each page. On a long statement, a helper column with =ISNUMBER(C5) (TRUE where the amount column holds a number) finds the records; Multi-page table to Excel shows the filter.
2. Mend wrapped descriptions. A description that wrapped onto a second line in the PDF arrives as a row of its own with no amount. Join its text to the row above and delete it.
3. Make the date what the import wants. Add a column with =DATE(RIGHT(A2,4),MID(A2,4,2),LEFT(A2,2)) for 09.10.2026, then format it: select the column, Ctrl+1, Custom, and type the import’s format, such as yyyy-mm-dd. Delete the original date column once the new one is checked.
4. Amounts. If the import wants 2 decimals, format the column as 0.00. If it wants separate debit and credit columns: =IF(C2<0,-C2,"") for debit and =IF(C2>0,C2,"") for credit.
5. Order and name the columns exactly as the import’s column map lists them, with the header row it expects.

Save it right: CSV UTF-8

Use File, Save As (or Save a Copy), and pick CSV UTF-8 (Comma delimited) as the file type, as Microsoft’s help “Save a worksheet as a CSV file” describes (checked on 11 October 2026). We tested what each choice writes, on a row with the name Jana Čermák:

Date,Description,Amount,Booking date
09.10.2026,"Invoice 2026-0412, Jana Čermák",1286.50,2026-10-09
14.10.2026,Refund,-1250.00,2026-10-14

• CSV UTF-8 kept Čermák intact and started the file with the 3 bytes of a UTF-8 marker (a byte order mark), so Excel recognises the encoding when it opens the file again. The converter’s own CSV starts with the same marker.
• CSV (Comma delimited), the older option, wrote “?ermák” with the á in an old Windows encoding. That damage is in the file and no import undoes it.
• Excel writes each cell as it is displayed: with the Amount column formatted 0.00, 1286.5 was saved as 1286.50; with the date formatted yyyy-mm-dd, the date was saved as 2026-10-09.
• A description containing the separator, like the comma in “Invoice 2026-0412, Jana Čermák”, is put in quotes, as the CSV standard requires.

The separator. Excel takes the list separator from your Windows region settings, not from the menu name; Microsoft’s help “Import or export text (.txt or .csv) files” (checked on 11 October 2026) explains how to change it. So always look at the first line of the saved file. If the import wants semicolons and the file has commas, the converter’s direct CSV with “Semicolon, 1190,50” is an alternative for simple lists, but note that it also writes a decimal comma.

Check before you import

Open the CSV in Notepad (Windows) or TextEdit (Mac), not in Excel, which would reinterpret it:

1. The first line is the header, with the separator you expect between the names.
2. The first and last record match the first and last transaction in the PDF.
3. The row count: the number of lines minus 1 equals the number of transactions. Notepad’s status bar shows the line number of the line you click.
4. Leading zeros in account or reference numbers are still there. If Excel removed them before you saved, the leading zeros guide puts them back.

With these 4 checks done, an import that still fails points at a column or a format, not at the whole file, and its error message tells you which.

The honest part

Every accounting import has its own specification: column order, date format, sign convention, separator. This guide gets the file clean; the final column map comes from your software’s import page, and the first import from a new bank or supplier deserves a careful look in the software afterwards. The converter does not know which rows are records, so the tidy pass is yours, and it does not read scanned PDFs: run OCR PDF first.

If your bank offers a CSV or other transaction export, that file is the bank’s own data and beats any conversion. The same goes for e-invoices: a ZUGFeRD or XRechnung carries the invoice as data your software can import directly (E-invoices and business PDFs). Amount and date formats from European documents are covered in PDF to Excel number formats, and the statement routine end to end in Bank statement to Excel.